Design of smart video surveillance system for indoor and outdoor scenes

Evaluation of a Realistic Example of Information-Centric Network Metadata Management

Abstract: This paper presents the performance evaluation of a metadata database (DB) management method that uses realistic numeric examples for IoT Live Data. The method is proposed to reduce the handling costs of Live Data. Live Data are here defined as data that are typically continuously generated by IoT devices and have short lifetimes (e.g., 10 fps surveillance camera images). We have already proposed an evaluation model in which the high locality is significantly featured in Live Data usage. The previous evaluation results are obtained only from general parameter values in statistical distributions. To evaluate realistic situations, this paper assumes that the metadata of Live Data with high usefulness for sharing by many users/services would dominate all metadata. In particular, for such data, we use both surveillance camera images and social networking service contents. The median values and the expected values are set considering the surveillance camera's locality (defined as the average distance between a surveillance camera and the users of its camera images). As a result, the proposed method can reduce the DB update costs by 99.0% while the additional search costs are reduced by up to 27.8% compared with the conventional metadata management method. The additional search costs are negligible compared with the reduction in DB update costs, since the number of searches is much smaller than the number of DB updates with respect to the number of update/search epochs.

Face detection for efficient video-surveillance IoT based embedded system

Abstract: Video Surveillance systems are widely used in indoor and outdoor environments for prevention and security monitoring. Most of conventional video surveillance systems are designed to store huge amount of data which difficult efficient access to the data from remote locations due to bandwidth requirements. A smart surveillance system allows efficient data storage and flexible data access. In this document the design and development of an embedded system for intelligent video surveillance with IoT capabilities is presented. For this project, an OMRON biometric sensor with specific features for face, body and hand detection was used. Face detection provides a criterion for event detection and efficient data capture of the data. The information of interest can be retrieved from a smartphone through Telegram X app. The system was tested under different face conditions including variations of pose, partial occlusion and expression. The system was developed with specific and smart devices providing new and different designs, easily to connect and control for users, without forgetting the importance of security.

Realtime dehazing using colour uniformity principle

Abstract: Dehazing is an important process as it can significantly improve the performance of computer vision applications in outdoor environments. The two main requirements that an online dehazing system demands are low processing time and high visual range. The authors present a novel dehazing algorithm based on colour uniformity principle (CUP) which meets the desired requirements of a realtime implementation. Estimation of atmospheric scattering parameter and transmission map forms the key step in dehazing problem. At first, the authors use CUP to generate the transmission map and refine it further by Fast Guided Filter. They estimate the atmospheric scattering parameter with the help of the estimated transmission map. Experimental results show that the quality of dehazed output, produced in real-time using the proposed method, is comparable with the results achieved by the state of the art techniques. The proposed dehazing method produces reliable dehazed output in varying haze conditions, unlike current methods.

Superintendence Video Summarization

Abstract: Today, System comprised of Surveillance cameras has become very useful and important in the every field, Mostly in the security industry. Also, Many numbers of surveillance cameras get added to the networks of surveillance or system every year as need and importance of surveillance cameras is increasing day by day. Video recorded from these surveillance cameras are large in size which require huge amount of time for monitoring and large storage space. Hence, there is a need of video summarization which has become very prominent since the last ten years because of the huge amount of available digital video content [3]. An algorithm we used for video summarization typically takes surveillance video as an input and extract a set of important frames or key-frames which is useful to represent the entire video content which are effectively more concise as compared to the original input video and convey semantic meaning. So, Our proposed solution aims at selecting keyframes from the video based on two criteria i.e. each object should appear within the scope of frame and each object should be visually presentable and must be closer to each other so that it could only show the related activities for ex. Summarization of video captured from ATM room camera should only display the part where user is interacting with the machine. So such a key frames are then used in final summarization.

Video Surveillance System with Auto Informing Feature

Abstract: The present document represents a thorough study of the making of an efficient surveillance system along with a feature of automatically informing the owner about the suspicious movement. In this moving world, normally people are suffering from the availability of time, so if any crime has happened at the site, it will take many days of searching for finding the actual presence of criminals, and thus a good chance for those burglars to flee away to protect themselves. For making the task possible, chose Python as the weapon for this battle and used different efficient techniques like COCO dataset for getting labeled and annotated images, LabelImg for making the annotation set of images, TensorFlow, object detection API for object detection and faster RCNN for training as faster RCNN has shown the highest accuracy for the COCO dataset so far. The owner can be informed in two ways: Either send a message to him via mail or phone or call at the time of suspicious image capturing. Here, both of these cases are used: For mail, the task is done via SMTP and for phone calls Twilio is used which provides us registered phone no. and can make both outbound and inbound calls. After using all the mentioned things and making the model in a way described above, it was found that faster RCNN is much more accurate than the other conventional methods. The results have been very well as RCNN show 86.7% accuracy and 100% has come out with the informing module as there simply the mail will be sent to the one whose mail is given in the code and the same is for Twilio calling.

A model of saliency-based visual attention for rapid scene analysis

Abstract: A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail.

Single Image Haze Removal Using Dark Channel Prior

Abstract: In this paper, we propose a simple but effective image prior-dark channel prior to remove haze from a single input image. The dark channel prior is a kind of statistics of outdoor haze-free images. It is based on a key observation-most local patches in outdoor haze-free images contain some pixels whose intensity is very low in at least one color channel. Using this prior with the haze imaging model, we can directly estimate the thickness of the haze and recover a high-quality haze-free image. Results on a variety of hazy images demonstrate the power of the proposed prior. Moreover, a high-quality depth map can also be obtained as a byproduct of haze removal.

Non-local Image Dehazing

Abstract: Haze limits visibility and reduces image contrast in outdoor images. The degradation is different for every pixel and depends on the distance of the scene point from the camera. This dependency is expressed in the transmission coefficients, that control the scene attenuation and amount of haze in every pixel. Previous methods solve the single image dehazing problem using various patch-based priors. We, on the other hand, propose an algorithm based on a new, non-local prior. The algorithm relies on the assumption that colors of a haze-free image are well approximated by a few hundred distinct colors, that form tight clusters in RGB space. Our key observation is that pixels in a given cluster are often non-local, i.e., they are spread over the entire image plane and are located at different distances from the camera. In the presence of haze these varying distances translate to different transmission coefficients. Therefore, each color cluster in the clear image becomes a line in RGB space, that we term a haze-line. Using these haze-lines, our algorithm recovers both the distance map and the haze-free image. The algorithm is linear in the size of the image, deterministic and requires no training. It performs well on a wide variety of images and is competitive with other stateof-the-art methods.